The standing wave pattern in reconstructed anomalous SLP resembles the standing Southern Oscillation wave (top left, Figure 2b) while that in reconstructed anomalous ST resembles the global standing ENSO wave pattern (top right, Figure 2b). This is also true of the propagating waves in Figure 2b and Figure 3b. This is rather remarkable since biennial signals are clearly separated from ENSO signals in zonal wavenumber frequency spectra (Figure 2a and Figures 3a). The propagating global biennial wave takes 2 to 3 years to circle the globe.
5. Quasi-Periodicity and Global Symmetries of Interdecadal Variability in Covarying SST and SLP anomalies.
In both White and Cayan (1999) and White and Allan (1999) the intensities of the global ENSO wave and the global biennial wave, as well as of the QBO and the ENSO, were found modulated by interdecadal variability. This is the dominant issue facing understanding of biennial and interannual variability; that is, how does interdecadal variability modulate biennial and interannual variability?
To begin addressing this issue, White and Cayan (1998) described the evolutionary patterns of interdecadal variability in both SLP and upper ocean temperature, finding it dominated by equatorward propagating wave activity in each basin, from high latitude to the tropics. In wavenumber-frequency spectra (Figure 1) we have already established the quasi-periodicity of interdecadal variability. The global symmetries of interdecadal variability are revealed in Figure 4a by exposing the 90-year record of interdecadal-filtered GISST SST and NCAR winter SLP anomalies to EOF analysis [Preisendorfer and Mobley, 1988]. However, they are not complete because of the paucity of data in the southern hemisphere during the first half of the 90-year period from 1900 to 1989. Therefore, we supplement this in Figure 4b with a similar EOF analysis applied to interdecadal-filtered GISST SST anomalies and SIO winter SLP anomalies for the 40 years from 1955 to 1994, extending from 40｡? to 60｡?.
The first two empirical orthogonal function (EOF) modes for both SST and winter SLP anomalies (Figure 4a) explain nearly equal amounts of interdecadal variance for a total of 45% for SST and 66% for SLP, forming quadrature pairs similar to those of the complex EOF [Preisendorfer and Mobley, 1988]. This indicates that global interdecadal variability is not characterized by standing wave modes over the globe but by propagating waves that reflect the time-space evolution of interdecadal variability examined in the next section. Amplitude time sequence of the dominant EOF modes for SST and SLP (Figure 4a, top) are approximately in phase over the 90-year record, even though they were computed independently from one another, with a tendency for the interdecadal anomaly pattern in SLP to lead that in SST by 2 years or so. They both display peaks near 1900, 1920, 1940, 1960, and 1980, with quasi-periodicity ranging from 20 to 22 years, much narrower than the 15 to 30 year period admittance window of the band pass filter. This time sequence is consistent with that associated with covarying patterns of northern hemisphere surface temperature and SLP on interdecadal timescales obtained by Mann and Park [1994].
The corresponding spatial pattern of the EOF mode in SST (Figure 4a, middle) displays positive weights for SST all across the global tropical ocean from approximately 20°S to 20°N, extending poleward into the extratropical eastern ocean in each basin with weights of opposite sign occurring along the western-central subarctic frontal zone (SAFZ) between 30° and 50°N in the northern hemisphere and between 25° and 40°S in the southern hemisphere.
